ML implements feed-forward artificial neural networks or, more particularly, multi-layer perceptrons (MLP), the most commonly used type of neural networks. MLP consists of the input layer, output layer, and one or more hidden layers. Each layer of MLP includes one or more neurons directionally linked with the neurons from the previous and the next layer. The example below represents a 3-layer perceptron with three inputs, two outputs, and the hidden layer including five neurons:
All the neurons in MLP are similar. Each of them has several input links (it takes the output values from several neurons in the previous layer as input) and several output links (it passes the response to several neurons in the next layer). The values retrieved from the previous layer are summed up with certain weights, individual for each neuron, plus the bias term. The sum is transformed using the activation function
..[LeCun98] Y. LeCun, L. Bottou, G.B. Orr and K.-R. Muller, *Efficient backprop*, in Neural Networks---Tricks of the Trade, Springer Lecture Notes in Computer Sciences 1524, pp.5-50, 1998.
..[RPROP93] M. Riedmiller and H. Braun, *A Direct Adaptive Method for Faster Backpropagation Learning: The RPROP Algorithm*, Proc. ICNN, San Francisco (1993).
Parameters of the MLP and of the training algorithm. You can initialize the structure by a constructor or the individual parameters can be adjusted after the structure is created.
The number of elements in each layer of network. The very first element specifies the number of elements in the input layer. The last element - number of elements in the output layer.
If you are using the default ``ANN_MLP::SIGMOID_SYM`` activation function with the default parameter values fparam1=0 and fparam2=0 then the function used is y = 1.7159*tanh(2/3 * x), so the output will range from [-1.7159, 1.7159], instead of [0,1].
Strength of the momentum term (the difference between weights on the 2 previous iterations). This parameter provides some inertia to smooth the random fluctuations of the weights. It can vary from 0 (the feature is disabled) to 1 and beyond. The value 0.1 or so is good enough
:param activateFunc:Parameter specifying the activation function for each neuron: one of ``ANN_MLP::IDENTITY``, ``ANN_MLP::SIGMOID_SYM``, and ``ANN_MLP::GAUSSIAN``.
:param fparam2:The second parameter of the activation function, :math:`\beta`. See the formulas in the introduction section.
:param termCrit:Termination criteria of the training algorithm. You can specify the maximum number of iterations (``maxCount``) and/or how much the error could change between the iterations to make the algorithm continue (``epsilon``).
Unlike many other models in ML that are constructed and trained at once, in the MLP model these steps are separated. First, a network with the specified topology is created using the non-default constructor or the method :ocv:func:`ANN_MLP::create`. All the weights are set to zeros. Then, the network is trained using a set of input and output vectors. The training procedure can be repeated more than once, that is, the weights can be adjusted based on the new training data.
Use ``StatModel::train`` to train the model, ``StatModel::train<ANN_MLP>(traindata, params)`` to create and train the model, ``StatModel::load<ANN_MLP>(filename)`` to load the pre-trained model. Note that the train method has optional flags, and the following flags are handled by ``ANN_MLP``:
***UPDATE_WEIGHTS** Algorithm updates the network weights, rather than computes them from scratch. In the latter case the weights are initialized using the Nguyen-Widrow algorithm.
***NO_INPUT_SCALE** Algorithm does not normalize the input vectors. If this flag is not set, the training algorithm normalizes each input feature independently, shifting its mean value to 0 and making the standard deviation equal to 1. If the network is assumed to be updated frequently, the new training data could be much different from original one. In this case, you should take care of proper normalization.
***NO_OUTPUT_SCALE** Algorithm does not normalize the output vectors. If the flag is not set, the training algorithm normalizes each output feature independently, by transforming it to the certain range depending on the used activation function.
